Search code examples
rcluster-analysisk-meanscovariance

Trying to modify the covariance matrix used in the inbuilt kmeans function in r

I am trying to locate where the kmeans function creates and assesses the covariance matrix from the input data set x. I want to override it and give it my own covariance matrix, because the unmodified one struggles with negative eigenvalues.

Another fix would be to find a function that does kmeans clustering with a modified covariance matrix as an inbuilt function, but I haven't found any such function yet.

Thanks in advance!

Function in question:

(stat:::kmeans)
Solution

  • KMeans uses neither a covariance matrix, nor a distance matrix, nor eigenvalues.

    On a proper distance matrix, negative eigenvalues cannot occur.

    You must be referring to some other algorithm. Reread kmeans - no covariance matrix, no eigenvalues.